Emotional experience in sheep: predictability of a sudden event lowers subsequent emotional responses

2 (2007) 675–683

Emotional experience in sheep: Predictability of a sudden event lowerssubsequent emotional responses

Lucile Greiveldinger ⁎, Isabelle Veissier, Alain Boissy

INRA, UR1213 Herbivores, Centre de Clermont-Ferrand/Theix, Adaptation et Comportements Sociaux, F-63122 Saint-Genès Champanelle, France

Received 24 October 2006; received in revised form 29 March 2007; accepted 10 May 2007

Abstract

The study of emotions in animals can be approached thanks to a framework derived from appraisal theories developed in cognitive psychology,according to which emotions are triggered when the individual evaluates challenging events. This evaluation is based on a limited number ofcriteria such as the familiarity and the predictability of an event. If animals are able to experience emotions rather than simply displaying reflexresponses to their environment, then their appraisal of events should, as in humans, modulate their emotional responses. We tested this hypothesisby comparing vocalisations, feeding behaviour, and the startle and cardiac responses of lambs submitted to a sudden event that could or could notbe predicted. Lambs able to predict the sudden event thanks to a light cue (associative predictability) showed weaker suddenness-induced startleand cardiac responses and spent more time feeding than their counterparts, thus supporting the existence of an emotional experience in theseanimals. Furthermore, lambs submitted to the regular appearance of the sudden event (temporal regularity) vocalised less and left less unconsumedfood deliveries than lambs submitted to random appearances of the sudden event (controls). These results underline that the cognitive abilities ofanimals should be taken into account when assessing their emotional experiences and more generally their mood states, which are underlyingfactors of animal welfare.© 2007 Elsevier Inc. All rights reserved.

Keywords: Predictability; Suddenness; Startle; Cardiac activity; Appraisal theories; Emotions; Animal welfare; Sheep

1. Introduction

Concern for the welfare of animals living in contact withhumans stems from the awareness that animals are sentientbeings potentially capable of experiencing emotions and moods[1–3]. However, it is unclear which emotions an animal is ableto experience. Dawkins [4] claimed that in order to understandanimal welfare, it was essential to consider “the animals' view-point”, which constitutes the link between physiological andbehavioural responses — which can be measured — and thesubjective experiences of the animal. But how can we assess theanimals' view-point without falling into the trap of anthropo-morphism? One efficient strategy is to use heuristic frameworks[5] that allow us to make unambiguous assumptions instead ofcase-by-case a posteriori interpretations of animals' responses.

It has been proposed that appraisal theories developed incognitive psychology [6–8] may provide a suitable framework for

⁎ Corresponding author. Tel.: +33 473 62 42 98; fax: +33 473 62 41 18.E-mail address: [email protected] (L. Greiveldinger).

0031-9384/$ - see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.physbeh.2007.05.012

studying emotions in animals. These theories suggest that emotionsstem from an individual's evaluation of challenging events [9], andthat this evaluation is processed via a limited number of criteria: theintrinsic characteristics of the situation (i.e. its suddenness,familiarity, pleasantness and predictability), its significance inrelation to the individual's needs (including conformity withexpectations), the coping potential available to the individual, andhow the situation fits with social and self-developed norms [10]. Ifanimals genuinely experience emotions rather than simply display-ing reflex responses, they can be expected to apply similar appraisalprocesses. In fact, appraisal, or at least intuitive appraisal, isnecessary for emotion to occur in humans [11]. However, appraisaldoes not imply conscious feelings. Our use of the term “emotionalexperience” refers to the basic experience of raw feelings alsoreferred to as “qualia” or “raw feels” [12], which requires aninformation process that distinguishes it from a simple reflex.

Extensive research has confirmed that mammals and birds dodisplay emotional responses (i.e. behavioural and physiologicalresponses) to sudden and unfamiliar events [13]. For instance,sheep facing a given physical event show variations in the nature

676 L. Greiveldinger et al. / Physiology & Behavior 92 (2007) 675–683

of their emotional responses according to whether the event issudden vs. unfamiliar [14]. This result further supports therelevance of the framework proposed by the appraisal theoriesfor assessing emotional experiences in animals. However, it canstill be argued that suddenness and unfamiliarity provokespecific reflex responses.

The question that remains is whether cognitive processessuch as those involved when animals make predictions oranticipations on their environment have an effect on the animal'semotional responses. Unpredictability appears to have negativeeffects on both humans and non-human animals: subjects optless for unpredictability than predictability in choice situations[15,16], and unpredictability can enhance negative emotionalexperiences [17,18], negative cognitive bias [19], fear [20],affected memory [21], anxiety [22,23], depression [24] or evenneurosis [25]. Some authors have even suggested thatbehavioural responses to unpredictability should be factoredinto the assessment of animal welfare [26].

Like most farm species, sheep are considered to have goodcognitive abilities, as shown from their performances in condi-tioning paradigms [27]. In the present study, we investigatewhether predictability affects emotional responses in sheep bycomparing their behavioural and physiological responses to asudden event they could or could not predict. Sudden events areknown to provoke in sheep a startle response together withtransient tachycardia [14]. We hypothesised that sheep wouldshow weaker emotional responses to a predictable sudden eventthan to an unpredictable sudden event.

2. Method

2.1. Animals

We used 36 five-month-old female lambs of the breed INRA401. The lambs were separated from their dam 12 h after birth inorder to reduce fear reactions to humans [28]. They were thenaccommodated in groups of three lambs in order to fulfil their socialneeds [29] and fed a milk substitute. After weaning (at 7 weeks ofage), all 36 lambs were reared indoors and housed together in asingle pen (7×7 m). The floor of the pen was bedded with woodchips and the room received ample lighting through largewindows.Room temperature was maintained at a constant 14 °C. Each lambwas given hay ad libitum plus 500 g of food pellets composed ofbarley, sugar beet, wheat, maize, sunflower and soya (ThivatNutrition Animale, Cusset, France) distributed daily at 5 pm.

2.2. Experimental set-up

The experimental chamber adjoining the home pen wascomposed of two pens (a pre-test and a post-test pen; each1.5×1.25 m) adjacent to the test arena (3×2.5 m) fenced in by1.8 m-high wooden barriers. Sliding doors permitted access fromthe home pen to the pre- and post-test pens and then to the testarena. The test arena contained a food pellet delivery device (Fig. 1)comprised of a deck placed at 30 cm above floor level and piercedwith a centre hole housing an adjustable 15 cm-diameter trough.Food pellets were distributed to this trough via an automatic system

placed outside the arena.A blue-and-white plastic panel was held at1 m above the trough behind a wooden board (i.e. hidden upposition). A second automatic device operated a sudden verticalmovement of the panel from the hidden up position to a visible lowposition just behind the trough (Fig. 1; speed about 2 m s−1). Twolights were fitted either side of the trough that could be switched onat the same time as the food delivery. Four cameras were used torecord behaviour: one camera filmed the whole experimentalchamber, while the three other cameras were placed one at eachside and one above the trough. All cameras were connected to avideo recorder and a monitor using a quadravision system giving asimultaneous recording of the four views.

2.3. Device to record cardiac activity

Cardiac activity was recorded via two adhesive electrodesplaced one on the right shoulder and the other on the left axillaryregion of the lamb. The electrodes were connected to a transmitterfixed to an elastic belt strapped around the animal's thorax (Fig. 1).A receiver (Lifescope 6, Nikon Kodhen, Japan) installed outsidethe experimental chamber received the signal by telemetry. Thisreceiver was connected to a Macintosh computer running a dataprocessing system (PowerLab, ADInstruments, GB). Cardiac datawas recorded and analysed using Chart software (version 3.6.8,HRVextension, ADInstruments, Australia).

2.4. Habituation procedure

When the lambs reached 15 weeks old, they were submitted toa habituation procedure lasting 4 weeks. FromDay 1 to Day 5, anexperimenter stayed in the home pen for 2 h a day (1 hour in themorning and 1 h in the afternoon). Between Day 6 and Day 10,the lambs were caught and an elastic belt was strapped aroundthe thorax twice a day for a 30-min period. OnDays 11 and 12, allthe doors were opened between the home pen, the pre- and post-test pens and the test arena to allow the lambs to freely explore allthe pens. FromDay 13 until the end of the experiment, each lambwas exposed once a day to the experimental chamber (i.e. onesession per day). From Day 13 to Day 19, each lamb was ledindividually to the pre-test pen where an experimenter fitted theelastic belt and two electrodes within about 30 s. The lamb wasthen left alone in the pre-test pen for another 30 s, after which thedoor to the test arena was opened and the lamb could go to thetrough containing 100 g of food pellets. When the lamb had eatenfor 30 s or after 5 min if it didn't eat, it was led to the post-test penwhere the belt and the electrodes were removed. The lambs couldthen return to the home pen. From Day 20 to Day 25, each lambreceived, in the test arena, 10 deliveries of 25 g food pellets in5 min, at a rate of one delivery every 30 s. The first delivery wasalready in the trough when the lamb entered the test arena. Foodpellets that were not consumed remained in the trough.

2.5. Experimental procedure

From Day 26 to Day 35 the experimental treatments wereapplied for 2 weeks with five sessions a week called test sessions.The procedures for handling and food delivery were the same as

Fig. 1. At each delivery, 25 g food pellets fell from the tube to the trough. Food deliveries were either followed (top right) or not (top left) by a sudden event generatedby a white and blue panel dropping down behind the trough from a hidden high point. Two round lights on either side of the tube could be turned on before theoccurrence of the sudden event (bottom). Before being introduced into the experimental chamber, the lamb was fitted with an elastic belt strapped around the thoraxthat held two electrodes and an emitter to measure and record heart rate (bottom).

677L. Greiveldinger et al. / Physiology & Behavior 92 (2007) 675–683

for the habituation procedure. Among the 10 food deliveries, fivewere followed by the sudden appearance of the plastic panel,called the sudden event, whichwas programmed 3 s after the lambhad started to eat.

The lambs were randomly allocated to one of three treatments(12 lambs per treatment): an associative predictability treatment, atemporal regularity treatment, and a control treatment. For theassociative predictability treatment, the schedule of food deliveryfollowed by a sudden event was chosen at random, but the lightswere turned on 6 s before the sudden event, at the same time as thefood delivery, for a total of 10 s. For the temporal regularitytreatment, the sudden event occurred after every other fooddelivery. For the control treatment (unpredictability), the fooddeliveries were followed by a sudden event that occurred randomly.

2.6. Measures

For each session, number of vocalisations was noted as ageneral indicator of stress, and the other behaviourswere analysed

based on the video recordings using The Observer software(Noldus, NL). We focused on feeding behaviour (i.e. number offood deliveries eaten and the percentage of time spent eating) asthe motivating factor, and on startle response (i.e. visible transientcontraction of the shoulder and/or hindquarter of the animaloccurring with a flexion of the legs or with the legs moving awayfrom each other) which is known to be a key behavioural responseto suddenness [14].

Heart rate, mean heart period (mean RR) and the root of themean squares of successive differences (RMSSD), which is anindicator of heart rate variability, were recorded. Taken together,these measurements make it possible to assess sympathovagalbalance [30].

2.7. Statistical analyses

Vocalisations and number of food deliveries eaten wereanalysed on the last session of the habituation period (Session-1,on Day 25) and on each of the ten test sessions (Sessions 1 to 10,

Fig. 2. Changes in the proportion of the 10 food deliveries completely, partially or not consumed over the last session of habituation (Session-1) and the test sessions(Sessions 1 to 9). We compared lambs given food deliveries followed by a sudden event occurring at random (control, n=12), lambs for which the sudden eventoccurred after every other food delivery (temporal regularity, n=12), and lambs for which the sudden event was preceded by a light signal (associative predictability,n=12).

678 L. Greiveldinger et al. / Physiology & Behavior 92 (2007) 675–683

from Day 26 to Day 35). Startle responses and percentage oftime spent eating were analysed for Session 1 (i.e. at the verybeginning of testing), Session 2 (after 1 day of testing), Session5 (after 4 days of testing) and Session 9 (after 8 days of testing).Finally, measurements of cardiac rate were analysed forSessions 1 and 5. Variables with two bounds (proportion ofdeliveries eaten, percentage of time spent eating) weretransformed into arcsine. Variables with one bound (cardiacvariables, number of startle responses) were log-transformed.First, to analyse changes across sessions, we analysed the 5 minof the test for all behavioural variables, for the last habituationsession and the ten test sessions. We ran ANOVAs for repeatedmeasures including session (i.e. the repetition), treatment and

Fig. 3. Mean and standard errors of time spent eating before and after the time the sucompared lambs given food deliveries followed by a sudden event occurring at randofood delivery (temporal regularity, n=12), and lambs for which the sudden event wassuperscript letters indicate significant difference (pb0.05).

their interaction. Second, we focused on the intervals immedi-ately preceding and following the sudden event in order toinvestigate the modulatory effect of predictability via anticipa-tion of and response to the sudden event. In that case, we ranANOVAs for repeated measures including session (i.e. therepetition, in that case only test sessions), treatment, occurrenceof the event (yes or no), interval (with two intervals except forthe cardiac data with three intervals: i) the interval between thefood delivery and the time of the possible sudden event, that is3 s after the animal had started eating (6.26 s±1.78 after fooddelivery); ii) the interval covering the following 10 s; and iii)only for cardiac data, the interval covering the 10 s preceding thefood delivery), and interactions between all factors.

dden event was due, according to whether the sudden event occurred or not. Wem (control, n=12), lambs for which the sudden event occurred after every otherpreceded by a light signal (associative predictability, n=12). Bars with different

Fig. 4. Changes in the mean and standard errors of the number of vocalisationsover the last session of habituation (Session-1) and the test sessions (Sessions 1to 9). We compared lambs given food deliveries followed by a sudden eventoccurring at random (control, n=12), lambs for which the sudden eventoccurred after every other food delivery (temporal regularity, n=12), and lambsfor which the sudden event was preceded by a light signal (associativepredictability, n=12).

Fig. 5. Mean and standard errors of startle responses at occurrence of the suddenevent. The data were pooled for Sessions 1, 2, 5 and 9. We compared lambsgiven food deliveries followed by a sudden event occurring at random (control,n=12), lambs for which the sudden event occurred after every other fooddelivery (temporal regularity, n=12), and lambs for which the sudden event waspreceded by a light signal (associative predictability, n=12). Bars with differentsuperscript letters indicate significant difference (pb0.05).

679L. Greiveldinger et al. / Physiology & Behavior 92 (2007) 675–683

We checked the conditions of the homogeneity of residualvariance. Post-hoc comparisons were performed with LeastSquare Means Differences (LSD). The Satterthwaite methodwas used for the degree of freedom estimation. The significancelimit was set at p=0.05.

3. Results

3.1. Food consumption

The number of food deliveries eaten decreased between the lastsession of habituation and the first test session, and increasedprogressively thereafter (Fig. 2; F10,258 (session) =8.17;pb0.001). The increasewasmore rapid in lambs in the associativepredictability treatment — with no significant increase after thesecond test session — than lambs in the control and temporalregularity treatments—with no significant increase after Session5 or Session 6 (Fig. 2; F20,258 (session× treatment)=1.75;p=0.03). Across the ten test sessions, the lambs in the controlgroup consumed less complete food deliveries (6.2 deliveries±2.1; F2,50 (treatment)=6.60; pb0.01) than lambs in treatmentgroups (8.1±2.2 for the temporal regularity treatment and 8.9±1.2 for the associative predictability treatment).

Time spent eating increased across sessions (Fig. 3; F3,2492

(session)=188.26; pb0.001). There were significant interac-tions between session, occurrence of the sudden event, andtreatment (Fig. 3; F18,2491 (session× treatment×moment×oc-currence of the event)=5.75; pb0.001). During the intervalfollowing food delivery, lambs in control and temporalregularity treatments showed no variation in the percentage oftime spent eating before the possible sudden event, whereaslambs in the associative predictability treatment group spent

proportionately more time eating when the sudden event wouldnot occur, and this pattern was visible as from the first testsession. From the second test session, the lambs from thecontrol and temporal regularity treatments spent less percentageof time eating during the interval following food delivery thanduring the interval following the sudden event time point whenthat sudden event did not occur. The same pattern was observedin lambs from the associative predictability treatment duringSession 2, but during Sessions 1, 5 and 9, they spent the sameamount of time eating during both intervals. For the intervalfollowing the sudden event when it did occur, during the firsttest session there were no between-treatment differences in theamount of time spent eating, but time spent eating thenincreased from Session 2 onwards in lambs of the associativepredictability treatment and from Session 5 onwards in the twoother treatment groups. Moreover, from the second test sessiononwards, lambs from the control treatment spent less timeeating following the sudden event than lambs in the other twotreatments groups.

3.2. Vocalisations

The frequency of vocalization increased markedly betweenthe last session of habituation and the first and second testsessions, and decreased progressively thereafter (Fig. 4; F10,285

(session)=5.68; pb0.001). The increase on the first and secondtest sessions and the subsequent decrease were less marked forlambs in the associative predictability group than in the othertwo treatments (F20,285 (treatment×session)=1.63; p=0.04).

3.3. Startle responses

The lambs demonstrated a startle response to the suddenevent (0.78 startle responses±0.01 after the sudden event vs.0.08±0.01 before the sudden event, F1,2488 (occurrence of theevent×moment)=1047; pb0.0001). These responses were less

Fig. 7. Mean and standard errors of the increase in heart rate between the 10 sintervals preceding and following the sudden event. The data were pooled forSessions 1 and 5. We compared lambs given food deliveries followed by asudden event occurring at random (control, n=12), lambs for which the suddenevent occurred after every other food delivery (temporal regularity, n=12), andlambs for which the sudden event was preceded by a light signal (associativepredictability, n=12). Bars with different superscript letters indicate significantdifference (pb0.05).

680 L. Greiveldinger et al. / Physiology & Behavior 92 (2007) 675–683

frequent in associative predictability-group lambs than in theother groups (Fig. 5; F2,2488 (treatment×occurrence of theevent×moment)=34.3; pb0.0001). During the interval be-tween the food delivery and the occurrence of the sudden event,there was no difference in the frequency of startle responsesbetween the associative predictability group with the light onand the two other treatments without the light (F2,2488

(treatment × occurrence of the event ×moment) = 34.3;pb0.0001). Across the ten test sessions, frequency of startleresponses remained stable in control lambs but decreased fromthe second test session in associative predictability-group lambsand from the ninth test session in temporal regularity-grouplambs (Fig. 6; F6,2495 (treatment×session)=2.51; pb0.02).

3.4. Cardiac activity

There were no differences in RMSSD between treatments,sessions and occurrences of the sudden event.

For all groups, Mean RR interval was lower during the fifthtest session than during the first test session (441 ms±12.5 vs.528 ms ± 12.5, respectively; F1,1807 (session) = 1248;pb0.0001). Before and after food delivery, there was nodifference in the Mean RR according to the treatment whetherthere was a sudden event or not (F6,1806 (treatment×occurrenceof the event×moment)=1.63; p=0.13). Mean RR decreased(i.e. the heart rate increased) just after the sudden event (Fig. 7;F2,1806 (occurrence of the event×moment)=32.5; pb0.0001).Mean RR in response to the sudden event was lower during thefirst test session than during the fifth test session (338 ms±13.0vs. 530 ms±13.0, respectively; F3,1806 (session×occurrence ofthe event×moment)=11.4; pb0.0001). On the fifth test session,mean RR was higher in the associative predictability-grouplambs than the control lambs after the time of the sudden event

Fig. 6. Changes in the mean and standard errors of number of startle responsesduring a 10 s interval following the sudden event. We compared lambs givenfood deliveries followed by a sudden event occurring at random (control,n=12), lambs for which the sudden event occurred after every other fooddelivery (temporal regularity, n=12), and lambs for which the sudden event waspreceded by a light signal (associative predictability, n=12). Bars with differentsuperscript letters indicate significant difference (pb0.05).

(respectively: 574 ms±22.1 vs. 493 ms±22.1; F4,1806 (session×treatment×moment)=2.59; p=0.04).

4. Discussion

This study questioned whether lambs' perception of theaversiveness of a sudden event varies with the predictability ofthat event. Behavioural and cardiac responses were measuredsimultaneously to first determine whether lambs could predict thesudden appearance of a plastic panel behind their trough based ona light signal cuing this event (associative predictability) or basedon its regularity over time (temporal regularity), and second todetermine whether the possibility to predict the event influencesthe subsequent emotional responses of the lambs.

The main results of this study are as follows: (i) Lambsdisplayed immediate responses to the sudden appearance of theobject in the form of a startle and a transient tachycardia; (ii)These responses were less marked when the appearance of theobject was preceded by the light signal; (iii) Lambs in the groupwhere the sudden appearance of the object was cued by the lightsignal spent more time eating when the light was not turned on;(iv) In the groups where there was no light signal cue, theresponses were more marked when the object appeared atrandom than when it appeared regularly over time.

The panel falling rapidly behind the trough triggered a startleresponse and a transient increase in heart rate in most lambs. Thisconfirms previous findings that lambs respond to sudden eventsby startle and tachycardia [14]. There were no variations inRMSSD. RMSSD reflects heart rate variability and is influencedby the parasympathetic branch of the autonomic nervous system[30]. Thus, the absence of variation in RMSSD suggests that theincrease in heart rate (and decrease in mean RR) arose mainlyfrom an activation of the sympathetic component of theautonomic nervous system rather than a lack of activation of theparasympathetic component. In the experiment reported here, thesudden event was probably perceived as aversive by the lambs:between the last habituation session (food reward with no sudden

681L. Greiveldinger et al. / Physiology & Behavior 92 (2007) 675–683

event) and the first test session where the sudden event followedfive food deliveries out of ten, we observed a significant decreasein food consumption and an increase in number of vocalisations.These behavioural patterns are classically considered to reflect anemotional reaction of animals exposed to stressful events [13].Nevertheless, over the ten test sessions, food consumptionincreased and the decrease in mean RR faded. Hence, the lambsappeared to habituate to the sudden appearance of the panel. Thenumber of startle responses decreased over the ten test sessions inthe temporal regularity and associative predictability groups butnot the control group (exposed to random occurrence of the eventnot cued by a light). Habituation to a sudden event is a well-known phenomenon [31,32], although it is difficult to ascertainwhether the decrease in response is due to habituation tosuddenness or to the event becoming less unfamiliar. Indeed, ithas been shown that the effects of suddenness and unfamiliarityare additive [33]. Changes over time are generally slower in thecase of repeated exposures to an unfamiliar and sudden event thanin the case of repeated exposures to an event that is unfamiliar butnot sudden. For instance, in our study, the sudden event triggeredthe same frequency of startle response in the control group fromthe first until the ninth test session, i.e. up to 45 occurrences of theevent, whereas lambs faced with an unfamiliar object have beenreported to stop walking backwards after only two to nineoccurrences [34]. Therefore, lambs habituate to a sudden event toa lesser extent than to simply an unfamiliar event.

Lambs for which the appearance of the panel was cued by alight signal spent less time eating just after food delivery when thelight was on than when it was off. This difference cannot beattributed to the light per se, since we have previously observedthat lambs spend more time eating in the presence of a light whenthe light signals the non-occurrence of the event rather than itsoccurrence (unpublished observations). The association of thelight with the appearance of the object may have led to a classicalconditioning paradigmwhere the conditioned response was “stopeating”. However, if this was the case, then the specific responseto the appearance of the panel would have been at least as high asthe response to the light. However, in our study, lambs spentproportionatelymore time eating during the interval following thepanel event than during the interval following the light signal.Moreover, if the light acted as a conditioned stimulus, then thelambs would have displayed tachycardia and a startle responsewhen the light was on. In our study, they did not display anytachycardia after the food deliveries, whatever the light was on oroff. Moreover they did not display more startle responses after thefood deliveries than their counterparts that were never submittedto the light. This suggests that the few startle responses observedafter the food deliveries were caused by the noise of the foodpellets falling in the trough instead of the switching on of the light.Our results suggest that lambs predicted that the panel wouldappear after the light signal and thus anticipated its occurrence bystopping eating (i.e. negative anticipation).

In response to the sudden appearance of the panel, lambsprovided with a light signal startled less often and showed lesstachycardia than lambs in the other groups. The magnitude ofthe response to a sudden event can decrease when the suddenevent is preceded by a weak sensorial event, e.g. a light flash

[35]. This phenomenon, called “pre-pulse inhibition” occurswhen the two events are separated by less than 500 ms, andoccurs on the first exposure to the combination of the sensorialand sudden events [36]. This phenomenon is not a likelyexplanation for our findings: the two events were separated byapproximately 6 s and several exposures were necessary toobserve a decrease in the response to the sudden event. It islikely that, in our experimentation, anticipation attenuated thesubsequent emotional responses to the event. The reduction instartle responses when an aversive event can be anticipated isconsistent with studies in humans [37–39]: the anticipation ofnegative events makes it possible to plan behavioural strategies.The majority of research on predictability in non-humananimals is focused on aversive events in rats [40]. In particular,Weiss [41] showed that the predictability of an electric shockmoderates its long-term negative effects (less gastric ulcers).Our results suggest that this long-term effect can at leastpartially due to animals experiencing the event as less aversiveat the exact time of its occurrence, thereby confirming the linkbetween the animal's perception of their environment (i.e. itscognitive appraisal) and subsequent emotional responses.

The lambs provided with a light cue before the occurrence ofthe panel emitted less vocalisations and showed a lower heartrate during the whole sessions than the lambs of the other groups.Vocalisations in farm animals are considered as an indicator ofnegative feelings [42,43]. Lambs that couldn't predict accuratelythe sudden events probably evaluated this sudden event andmore generally the whole test sessions as being more stressfulthan the lambs from the associative predictability treatment.

The lambs that were exposed to the sudden appearance of thepanel regularly, i.e. after every other food delivery, showed nodifference in pre- or post-event behaviour compared to lambsexposed to the event at random and without a signal. Thus, incontrast with our assumption, it appears that the lambs did notuse the food deliveries alternating with or without sudden eventto predict the sudden event. This suggests that lambs may be notequipped with representation of time that allows them to copewith events occurring in an alternating binary sequence. Ourlambs were maybe too young (6 months old) and/or this speciesmay not have the necessary cognitive capacity [44,45].Alternatively, our lambs may not have been trained for longenough to learn the alternating sequence. Another study usingtemporal regularity led to unclear results on whether calvescould predict an event based on a regular time schedule [46].Hence, young ruminants like calves or lambs seem able topredict an event when it is explicitly signalled but havedifficulties predicting it on the bases of regular intervals.

However, over whole test sessions, the lambs in the temporalregularity treatment group vocalised less and ate more than thecontrol-group lambs exposed to the panel on random occasions.The lambs in the temporal regularity group were always given afixed 1min interval between two sudden events (a food delivery notfollowed by the sudden event was always preceded and followedby a food delivery plus the sudden event, and vice versa). It hasbeen shown that the use of fixed-time schedule for aversive eventsis less stressful than a variable-time schedule [47]. Hence, lambs inthe temporal regularity treatment may have understood that there

682 L. Greiveldinger et al. / Physiology & Behavior 92 (2007) 675–683

was always a ‘safe’ few-second period just after a sudden event,even if they were not able to accurately predict the followingsudden event. An important concept in predictability studies is the“safety signal” hypothesis according to which the fear-reducingproperties of the predictability of a stressful event is more due to thepredictability of the safe inter-event periods than to thepredictability of the stressful event itself [40,47–51]. Thus, anindividual who has the ability to predict the periods withoutstressful event can relax during those periods. More specifically,Badia et al. [47] showed that all human subjects given a choicebetween fixed-time and variable-time unsignaled shocks chose thefixed-time schedule. Thus, if lambs in the temporal regularitytreatment were able to discriminate safe moments during thesessions even though they were unable to predict the exact time ofthe sudden event, they should have perceived the test sessions asless stressful than control lambs which were always aroused. Thisinterpretation is confirmed by the fact that behaviours reflectingunpleasantness (e.g. vocalisations and food refusals) disappearedfaster in temporal regularity-group lambs than in controls. Theseresults could have been strengthened by measuring pain inhibitionor cortisol production, which are shown to increase in response tounpredictability [52–55]. However, short-term studies usingcortisol measurements give inconsistent results [56].

The differences observed between the three treatments couldresult from interplay between different predictability levels ofthe sudden event and different levels of discrimination of reliablesafety periods. In this study, these reliable safety periods wereclearly identified by lambs in the associative predictability groupvia the absence of the light signal and were less clearly identifiedby lambs in the temporal regularity treatment as occurring duringthe few seconds immediately following the sudden event.

In conclusion, lambs evaluate the events occurring in theirenvironment according to their predictability, and this evalua-tion modulates their emotional responses. This study confirmsthe relevance of using appraisal theories to assess animals'emotions. Sheep reactions to their environment are not simplyreflex responses and therefore have to be considered as theresult of an appraisal process, similarly to human beings.Research on the brain structures involved in the modulation ofemotional responses should make it possible to highlight theevaluative processes influencing emotional perception andemotional responses in animals [57]. This study opens theway to further work on more complex cognitive appraisalcriteria, including the controllability of an event or deviationfrom expectations, and whether these criteria are suitable forassessing emotion in animals. From a practical view-point,increasing predictability in the environment of farmed orcaptive animals could potentially improve animal welfare[40,58], although some authors underline the risk that anenvironment that is too predictable may induce boredom [59].

Acknowledgments

The authors are most grateful to S. Andanson, E. Delval andG. Toporenko for their help in data recording and analysis, to C.Ravel and B. Mallet for animal care, and to T. Vimal, J. Ferreiraand J.P. Chaise for their technical support.

References

[1] Gonyou HW. Why the study of animal behavior is associated with theanimal welfare issue. J Anim Sci 1994;72:2171–7.

[2] Duncan IJH. Poultry welfare: science or subjectivity? Br Poult Sci2002;43:643–52.

[3] Mendl M, Paul ES. Consciousness, emotion and animal welfare: insightsfrom cognitive science. Anim Welf 2004;13:S17–25.

[4] Dawkins MS. From an animal point of view: motivation, fitness, andanimal welfare. Behav Brain Sci 1990;13:1–9.

[5] Wynne CDL. The perils of anthropomorphism. Nature 2004;428:606.[6] Dantzer R. Can farm animal welfare be understood without taking into

account the issues of emotion and cognition? J Anim Sci 2002;80:E1–9.[7] Désiré L, Boissy A, Veissier I. Emotions in farm animals: a new approach

to animal welfare in applied ethology. Behav Processes 2002;60:165–80.[8] Paul ES, Harding EJ, Mendl M. Measuring emotional processes in

animals: the utility of a cognitive approach. Neurosci Biobehav Rev2005;29:469–91.

[9] Lazarus RS, Averill JR, Opton Jr EM. Towards a cognitive theory ofemotion. In: Arnold MB, editor. Feelings and emotions: the LoyolaSymposiums. New York: Academic Press; 1970. p. 207–32.

[10] Scherer KR. Appraisal theory. In: Dalgleish T, Power M, editors.Handbook of cognition and emotion. Chichester: John Wiley & Sons;1999. p. 637–63.

[11] Kappas A. Appraisals are direct, immediate, intuitive, and unwitting... andsome are reflective.... Cogn Emot 2006;20:952–75.

[12] Dawkins MS. Through animal eyes: what behaviour tells us. Appl AnimBehav Sci 2006;100:4–10.

[13] Boissy A. Fear and fearfulness in determining behavior. In: Grandin T,editor. Genetics and the behavior of domestic animals. New York:Academic press; 1998. p. 67–111.

[14] Désiré L, Veissier I, Després G, Boissy A. On the way to assess emotionsin animals: do lambs evaluate an event through its suddenness, novelty orunpredictability? J Comp Psychol 2004;118:363–74.

[15] Mineka S, Hendersen RW. Controllability and predictability in acquiredmotivation. Annu Rev Psychol 1985;36:495–529.

[16] Badia P, Culbertson S, Harsh J. Choice of longer or stronger signalledshock over shorter or weaker unsignalled shock. J Exp Anal Behav1973;19:25–32.

[17] Adkin AL, Quant S, Maki BE, McIlroy WE. Cortical responses associatedwith predictable and unpredictable compensatory balance reactions. ExpBrain Res 2006;172:85–93.

[18] Carlsson K, Andersson J, Petrovic P, Petersson KM, Ohman A, Ingvar M.Predictability modulates the affective and sensory-discriminative neuralprocessing of pain. NeuroImage 2006;32(4):1804–14.

[19] Harding EJ, Paul ES, Mendl M. Cognitive bias and affective state. Nature2004;427:312.

[20] Armfield JM. Cognitive vulnerability: a model of the etiology of fear. ClinPsychol Rev 2006:746–68.

[21] Burger JM, Arkin RM. Prediction, control, and learned helplessness. J PersSoc Psychol 1980;38:482–91.

[22] ZvolenskyMJ, Georg HE, Lejuez CW, Hopko DR, Forsyth JP. Assessing theperceived predictability of anxiety-related events: a report on the perceivedpredictability index. J Behav Ther Exp Psychiatry 2000;31:201–18.

[23] Zvolensky MJ, Lejuez CW, Eifert GH. Prediction and control: operationaldefinitions for the experimental analysis of anxiety. Behav Res Ther2000;38:653–63.

[24] Anisman H, Matheson K. Stress, depression, and anhedonia: caveatsconcerning animal models. Neurosci Biobehav Rev 2005;29(4-5):525–46.

[25] Mineka S, Kihlstrom JF. Unpredictable and uncontrollable events: a newperspective on experimental neurosis. JAbnormPsychology 1978;87:256–71.

[26] SandemAI, Janczak AM, Braastad BO. A short note on effects of exposureto a novel stimulus (umbrella) on behaviour and percentage of eye-white incows. Appl Anim Behav Sci 2004;89:309–14.

[27] Nicol CJ. Farm animal cognition. Anim Sci 1996;62:375–91.[28] Romeyer A, Bouissou M-F. Assessment of fear reactions in domestic

sheep, and influence of breed and rearing conditions. Appl Anim BehavSci 1992;34:93–119.

683L. Greiveldinger et al. / Physiology & Behavior 92 (2007) 675–683

[29] Penning PD, Parsons AJ, Newman JA, Orr RJ, Harvey A. The effects ofgroup size on grazing time in sheep. Appl Anim Behav Sci1993;37:101–9.

[30] Després G, Boissy A, Désiré L, Le Neindre P, Veissier I. Validation of themeasure of sympatho-vagal effect in lambs through autonomic blockadesand heart rate variability indexes. J Anim Vet Adv 2003;2:615–9.

[31] Turpin G, Siddle DAT. Effects of stimulus intensity on cardiovascularactivity. Psychophysiology 1983;20:611–24.

[32] Groves PM, Thompson RF. Habituation: a dual-process theory. PsycholRev 1970;77:419–50.

[33] Désiré L, Veissier I, Després G, Delval E, Toporenko G, Boissy A.Appraisal process in sheep (Ovis aries): interactive effect of suddennessand unfamiliarity on cardiac and behavioral responses. J Comp Psychol2006;120(3):280–7.

[34] Désiré L. Etude des processus cognitifs impliqués dans la différenciationdes émotions chez l'agneau (Ovis aries). PhD thesis manuscript 2004.

[35] Hoffman HS, Ison JR. Reflex modification in the domain of startle: I.Some empirical findings and their implications for how the nervous systemprocesses sensory input. Psychol Rev 1980;87:175–89.

[36] Braff DL, Geyer MA, Swerdlow NR. Human studies of prepulse inhibitionof startle: normal subjects, patient groups, and pharmacological studies.Psychopharmacology 2001;156:234–58.

[37] Ekman P, Friesen WV, Simons RC. Is the startle reaction an emotion?J Pers Soc Psychol 1985;49:1416–26.

[38] Grillon C, Baas JP, Lissek S, Smith K, Milstein J. Anxious responses topredictable and unpredictable aversive events. Behav Neurosci2004;118:916–24.

[39] Hagemann T, Levenson RW, Gross JJ. Expressive suppression during anacoustic startle. Psychophysiology 2006;43:104–12.

[40] Bassett L, Buchanan-Smith HM. Effects of predictability on the welfare ofcaptive animals. Appl Anim Behav Sci 2007;102(3–4):223–45.

[41] Weiss JM. Psychological factors in stress and disease. Sci Am1972;226:104–13.

[42] Vandenheede M, Bouissou MF. Sex differences in fear reactions in sheep.Appl Anim Behav Sci 1993;37:39–55.

[43] Manteuffel G, Puppe B, Schön PC. Vocalization of farm animals as ameasure of welfare. Appl Anim Behav Sci 2004;88:163–82.

[44] Leventhal H, Scherer K. The relationship of emotion to cognition: afunctional approach to a semantic controversy. Cogn Emot 1987;1:3–28.

[45] Toates F. Cognition, motivation, emotion and action: a dynamic andvulnerable interdependence. Appl Anim Behav Sci 2004;86:173–204.

[46] Johannesson T, Ladewig J. The effect of irregular feeding times on thebehaviour and growth of dairy calves. ApplAnimBehav Sci 2000;69:103–11.

[47] Badia P, Harsh J, Coker CC. Choosing between fixed time and variabletime shock. Learn Motiv 1975;6:264–78.

[48] Matthews KA, Scheier MF, Brunson BI, Carducci B. Attention,unpredictability and reports of physical symptoms: eliminating the benefitsof predictability. J Pers Soc Psychol 1980;38:525–37.

[49] Grillon C, Ameli R. Effects of threat and safety signals on startle duringanticipation of aversive shocks, sounds, or airblasts. J Psychophysiol1998;12:329–37.

[50] Tsuda A, Tanaka M, Hirai H, Pare WP. Effect of coping behavior on gastriclesions in rats as a function of predictability of shock. Jpn Psychol Res1983;25:9–15.

[51] Weinberg J, Levine S. Psychobiology of coping in animals: the effects ofpredictability. In: Levine S, Ursin H, editors. Coping and health. NewYork: Plenum Press; 1980. p. 39–59.

[52] Muir JL, Pfister HP. Corticosterone and prolactin responses to predictableand unpredictable novelty stress in rats. Physiol Behav 1986;37:285–8.

[53] Harlow HJ, Thorne ET, Williams ES, Belden EL, Gern WA. Adrenalresponsiveness in domestic sheep (Ovis aries) to acute and chronicstressors as predicted by remote monitoring of cardiac frequency. Can JZool 1987;65:2021–7.

[54] Tazi A, Dantzer R, Le Moal M. Prediction and control of food rewardsmodulate endogenous pain inhibitory systems. Behav Brain Res1987;23:197–204.

[55] Pinto-Ribeiro F, Almeida A, Pêgo JM, Cerqueira J, Sousa N. Chronicunpredictable stress inhibits nociception in male rats. Neurosci Lett2004;359:73–6.

[56] Abbott BB, Schoen LS, Badia P. Predictable and unpredictable shock:behavioral measures of aversion and physiological measures of stress.Psychol Bull 1984;96:45–71.

[57] Hamm AO, Weike AI. The neuropsychology of fear learning and fearregulation. Int J Psychophysiol 2005;57:5–14.

[58] Bloomsmith MA, Lambeth SP. Effects of predictable versus unpredictablefeeding schedules on chimpanzee behavior. Appl Anim Behav Sci1995;44:65–74.

[59] Van Rooijen J. Predictability and boredom. Appl Anim Behav Sci1991;31:283–7.